Viktoria F.
Kunzelmann
,
Chang-Ming
Jiang
,
Irina
Ihrke
,
Elise
Sirotti
,
Tim
Rieth
,
Alex
Henning
,
Johanna
Eichhorn
and
Ian D.
Sharp
*
Walter Schottky Institute and Physics Department, Technische Universität München, Am Coulombwall 4, 85748 Garching, Germany. E-mail: sharp@wsi.tum.de
First published on 22nd April 2022
We demonstrate a facile approach to solution-based synthesis of wafer-scale epitaxial bismuth vanadate (BiVO4) thin films by spin-coating on yttria-stabilized zirconia. Epitaxial growth proceeds via solid-state transformation of initially formed polycrystalline films, driven by interface energy minimization. The (010)-oriented BiVO4 films are smooth and compact, possessing remarkably high structural quality across complete 2′′ wafers. Optical absorption is characterized by a sharp onset with a low sub-band gap response, confirming that the structural order of the films results in correspondingly high optoelectronic quality. This combination of structural and optoelectronic quality enables measurements that reveal a strong optical anisotropy of BiVO4, which leads to significantly increased in-plane optical constants near the fundamental band edge that are of particular importance for maximizing light harvesting in semiconductor photoanodes. Temperature-dependent transport measurements confirm a thermally activated hopping barrier of ∼570 meV, consistent with small electron polaron conduction. This simple approach for synthesis of high-quality epitaxial BiVO4, without the need for complex deposition equipment, enables a broadly accessible materials base to accelerate research aimed at understanding and optimizing photoelectrochemical energy conversion mechanisms.
In this work, we introduce a new approach for wafer-scale synthesis of epitaxial ms-BiVO4 on yttria-stabilized zirconia (YSZ) (001) using a solution-based process comprising spin-coating, metalorganic decomposition, and solid-state epitaxial transformation. This technique allows facile fabrication of thin films possessing a high optoelectronic quality and a structural quality that exceeds previously reported epitaxial BiVO4 films produced using much more sophisticated deposition approaches, as exemplified by high-resolution X-ray diffraction (HR-XRD) measurements. We find that, compared to polycrystalline thin films, (010) BiVO4 is characterized by significantly increased band-edge optical constants due to the strong uniaxial anisotropy of the material, indicating the importance of orientation control on maximizing light absorption in photoelectrodes. More generally, this facile and scalable method for synthesizing epitaxial BiVO4 will enable widespread availability of highly ordered material required for elucidating physical and chemical mechanisms of photoelectrochemical energy conversion.
The BiVO4 precursor solution was prepared in a 1:
1 ratio from two separate solutions, one containing 0.2 M bismuth(III) nitrate pentahydrate (Bi(NO3)3·5H2O, Sigma-Aldrich) in acetylacetone (C5H8O2, Sigma-Aldrich) and the other containing 0.03 M vanadium(IV)-oxy acetylacetonate (OV(C5H7O2)2, Sigma Aldrich) in acetylacetone. Each solution was sonicated for 10 min, after which the solutions were mixed and the final precursor solution was sonicated for an additional 5 min. Finally, the prepared solution was filtered with a PET membrane filter (0.2 μm pore size, 15 mm diameter, Carl Roth) to ensure a clean and fully mixed precursor. A drop of 500 μl precursor solution was spin-coated on a YSZ wafer with an acceleration rate of 150 rpm s−1, a spin speed of 1000 rpm for 12 s, and a deceleration rate of 150 rpm s−1. Subsequently, the wafer was placed on a 1 mm thick steel plate and annealed at 500 °C for 10 min in air at ambient pressure in a muffle furnace (LT 3/12, Nabertherm). After annealing, the wafer was taken out of the oven to cool down to room temperature for 12 min. Repeating this growth sequence for nine layers resulted in an approximately 43 nm thick polycrystalline (poly-)BiVO4/YSZ thin film. To achieve an epitaxial, single-phase BiVO4 layer, the resulting film was annealed at 650 °C for 10 min in a confined air volume of approximately 600 mm3. The volume confinement in the muffle furnace was formed by placing a clean, thermally stable silicon wafer approximately 300 μm above the BiVO4/YSZ wafer, supported by other wafer pieces that still allowed external air to enter. The purpose of this confinement was to mitigate loss of volatile elements from the surface during the annealing at 650 °C, thereby enabling a stoichiometric film to be retained. This configuration may have also reduced convective air flow over the surface, thereby improving temperature uniformity during the annealing treatment.
For reference measurements of polycrystalline BiVO4 on fused silica, wafers of fused silica (T24003, Siegert Wafer) with a diameter of 10 cm were rinsed with isopropanol and carefully scrubbed with a cleaning swab in a detergent solution (Alconox, Alconox Ing.). After rinsing with deionized water, the wafer was treated in an ozone cleaner for 10 min (UVC-1014, NanoBioAnalytics). The precursor preparation and spin-coating procedure were similar to the ones described above for the BiVO4/YSZ films, except that 1 ml precursor solution per layer was used, the wafer was placed on a 1 mm thick aluminum plate, and a final 2 h anneal was performed at 500 °C, as previously reported in the literature.20
Raman spectra were obtained with a home-built setup with a 532 nm wavelength laser (Torus 532, Laser Quantum). Prior to measurement, the system was calibrated to the 520 cm−1 peak of a single-crystal silicon reference. The backscattered light was analyzed with a spectrometer (iHR5500, HORIBA Scientific) equipped with a 2400 mm−1 grating and a CCD detector (Symphony II, HORIBA Scientific).
X-ray photoelectron spectroscopy data were recorded with a Kratos Axis Ultra setup equipped with a monochromatic Al Kα X-ray source and an applied power of 225 W. To avoid surface charging, charge neutralization was performed with the Kratos Axis electron charge neutralization system with an applied filament current of 0.45 A, a filament bias voltage of 1 V, and a charge balance voltage of 3 V. The survey spectra were measured with a pass energy of 160 eV and a step size of 1 eV, and the core level spectra were obtained with a pass energy of 10 eV and a step size of 0.03 eV. Remaining charging artifacts prevent detailed spectral fitting of individual components.
For the investigation of the bulk composition of the epitaxial BiVO4 films, a Bruker XFlash6130 energy-dispersive X-ray spectrometer, mounted in a Zeiss EVO MA10 for sample alignment imaging, was used. The EDX detection spot size was 1 μm and an electron beam energy of 20 keV was applied. The resulting spectra were analyzed using series fit deconvolution and the Phi-Rho-Z quantification mode.
A Bruker MultiMode 8 atomic force microscope was used for the topography measurements. All measurements were acquired in tapping mode with NSG30 tips (NT-MDT). The images were post-processed with the Bruker Nanoscope Analysis software with 0th order flattening and a 3rd order plane fit to account for instrumental aberrations. Prior to recording the high-resolution and cross-sectional scanning electron microscopy (SEM) images, a ∼3 nm thin carbon coating was sputtered on the investigated BiVO4 films to reduce charging. For the cross-sectional SEM images, the samples were then freshly cleaved to obtain a sharp edge. The SEM images were obtained with an NVision 40 FIB-SEM from Carl Zeiss using the secondary electron detector and an electron beam acceleration voltage of 2.0 kV.
(αhν)n = A(hν − Eg) |
The optical constants of the BiVO4 films on YSZ and on fused silica were measured with a variable-angle spectroscopic ellipsometer from J. A. Wollam under a light incidence angle of 55°, 60° and 65°. The refractive index, n, the extinction coefficient, κ (Fig. S4a and c†), and the real, ε1, and imaginary, ε2, parts of the dielectric constant (Fig. S4b and d†) were obtained by fitting the spectroscopic ellipsometry data with a layer model developed with the CompleteEASE software from J. A. Wollam. The YSZ substrate was fit with a B-spline model with a fixed thickness of 0.5 mm and a Bruggeman effective medium approximation surface layer accounting for surface roughness. The B-spline layer was then parametrized as a general oscillator (GenOsc) layer consisting of a Cody–Lorentz oscillator including an Urbach absorption term to model absorption at energies below Eg, a Drude oscillator describing free carrier effects on the dielectric response, and a Gaussian oscillator, all fulfilling Kramers–Kronig consistency. For the fused silica substrate, the model as described by Philipp22 and found in the CompleteEASE database was used. The BiVO4 film was modelled as an additional B-spline layer on the YSZ substrate which was parametrized by a GenOsc layer with a three oscillator model as described by Cooper et al.,6 with: a Cody–Lorentz oscillator for modelling the indirect band gap of BiVO4, including parameters describing the Urbach tail associated with sub-band gap absorption induced by disorder, a PSemi-M0 oscillator for modeling the direct electronic transition in the semiconductor, and a PSemi-Tri oscillator to describe the remaining UV absorption.6
Cleaning and thermal pre-treatment of the YSZ wafer were performed to achieve high wettability, coverage, and homogeneity during the spin-coating process (Fig. S1†). The synthetic route to epi-BiVO4 layers (for complete details see Methods) is similar to the metalorganic decomposition method that has been previously used for fabrication of polycrystalline material.20,26 In brief, a precursor solution was mixed from two solutions containing 0.2 M bismuth(III) nitrate pentahydrate and 0.03 M vanadium(IV)-oxy acetylacetonate, each dissolved in acetylacetone. As shown in Fig. 1a, the precursor solution was then spin-coated onto a two-inch (2′′) YSZ (001) wafer and subsequently pyrolyzed at 500 °C for 10 min in a muffle furnace in ambient air. Repeating this growth sequence for nine layers resulted in an approximately 43 nm thick BiVO4 film with a polycrystalline structure, which was confirmed by grazing-incidence XRD (GI-XRD) (Fig. S2a†). To transform the polycrystalline films into epi-BiVO4, we performed a final annealing treatment at 650 °C for 10 min in ambient air. Raman spectroscopy confirms that the films comprise the monoclinic scheelite polymorph of BiVO4 both before and after this annealing treatment (Fig. S6†), as indicated by the (VO4)3−ν1 symmetric stretching mode at ∼826 cm−1.27
We note that this last annealing step is performed at a much higher temperature and for a shorter time than for the typical metalorganic decomposition approach for forming poly-BiVO4, which is usually carried out at 450–500 °C for several hours. This short time, high-temperature step is crucial for inducing a solid-state transformation of poly-BiVO4 into an epitaxial thin film, which will be discussed in detail below. We note that an advantage of this two-step method is that the epitaxial transformation is independent of the specific chemical route used to prepare the initial polycrystalline layer. Therefore, we expect this process to be broadly compatible with the various precursor solution compositions and solvents that have been reported for wet chemical synthesis of polycrystalline BiVO4 films.
At the high annealing temperature of 650 °C, it is known that vanadium becomes volatile in air.28 Therefore, this step was performed in a confined volume, which helped to maintain the desired stoichiometry. Consistent with prior reports of metalorganic decomposition-derived BiVO4,20,29 X-ray photoelectron spectroscopy (XPS) reveals a Bi-rich surface (Fig. S7†). The surface Bi:
V ratio is found to be ∼1.7 for both the poly- and epi-BiVO4 films on YSZ, indicating that the combination of short time and spatial confinement during the 650 °C annealing step suppresses further volatilization of vanadium. Investigation of the bulk composition with energy-dispersive X-ray spectroscopy (EDX) revealed an atomic fraction of ∼48.4 at% bismuth and ∼51.6 at% vanadium (Fig. S8†). This bulk composition is consistent with the results of previous studies on similarly prepared polycrystalline thin films20 and confirms the 1
:
1 ratio between Bi and V in BiVO4 to within the few percent uncertainty of the EDX measurement.
A key advantage of spin-coating is its potential for wafer-scale growth. To assess the growth homogeneity, a systematic series of θ–2θ scans was performed across the entire 2′′ YSZ wafer. At each measurement position, comparable diffractograms were observed (Fig. 1d) and the FWHMs of the corresponding BiVO4 (040) rocking curves are between 0.035° and 0.05° (Fig. S2d†). These measurements confirm the successful growth of a homogeneous, high-quality epi-BiVO4 (010) film over the entire surface of the YSZ (001) wafer and could be reliably reproduced in several epi-BiVO4/YSZ wafer growth runs (Fig. S9†). In addition, we performed ellipsometry measurements across the diameter of the epitaxial BiVO4 film, confirming a uniform thickness of approximately 43 nm (Fig. S10†). Edge effects that are common to spin-coating resulted in film thickening, but were limited to a 4 mm width around the circumference of the wafer.
The in-plane epitaxial relationship between BiVO4 and the YSZ substrate is revealed by azimuthal φ-scans of YSZ (204) and BiVO4 (082) reflections. As shown in Fig. 1e, both reflections exhibit a 4-fold symmetry and are offset by a ∼45° angle. Given the BiVO4 unit cell definition with a unique b-axis adopted in this work, its (082) reflection is expected to have a 2-fold symmetry. Therefore, the observed 4-fold symmetry is indicative of 90° in-plane twinning, with domains characterized by in-plane relationships of BiVO4 [001] ‖ YSZ [100] and BiVO4 [001] ‖ YSZ [010] (Fig. S11†). This assignment is further supported by 2D reciprocal space maps (RSM) with X-rays incident along the YSZ [100] direction (Fig. 1f). Not only does the pattern match with the simulated BiVO4 [001](010) ‖ YSZ [100](001) model (Fig. S12a†), but the (330), (440), and (260) BiVO4 reflections each exhibit a doublet feature caused by the 90° in-plane twinning. Similar conclusions can be drawn from the RSMs with X-rays incident along the YSZ [110] and [010] directions, as shown in Fig. S12b and c.†
Here, it is important to recognize that BiVO4 adopts the more symmetric tetragonal scheelite structure at the temperature used for the conversion of polycrystalline to epitaxial films and undergoes a second-order ferroelastic transformation to the monoclinic structure upon cooling.30,31 Therefore, azimuthal twinning occurs during cooling, after the out-of-plane epitaxial relationship has been established, and multi-domain films are commonly observed for epitaxial BiVO4 (010) on YSZ.9,15,32 Although single domain epitaxial BiVO4 (010) films have been claimed by a few groups,10–12,14 the experimental evidence for single domain films remains limited.
Both AFM and SEM images reveal the presence of scattered particles, which is consistent with a small quantity of remnant mis-oriented grains detected by GI-XRD (Fig. S2a†). The occurrence of these grains points to the mechanism of the solid-state transformation from polycrystalline to epitaxial films, as described in the following. After the spin-coating and 500 °C pyrolysis steps, pole figure measurements indicate a polycrystalline but textured BiVO4 film with preferred (010) orientation (Fig. S14†). Such film characteristics are consistent with parallel heterogeneous and homogeneous nucleation events occurring during pyrolysis.33 In particular, it has been shown that moderate thermal ramp rates, such as those used here (Fig. S15†), promote nucleation over a broad range of relatively low temperatures, where driving forces for both homogeneous and heterogeneous nucleation are large, resulting in multiple grain configurations.33,34 Subsequent high-temperature annealing (650 °C) induces epitaxial grain coarsening via the consumption of mis-oriented grains, energetically driven by interface energy minimization and kinetically enabled by high diffusivities.35 Although the fine-grained morphology of the original polycrystalline film facilitates rapid grain boundary motion and epitaxial film formation, the driving force for this transformation decreases with decreasing fraction and size of mis-oriented grains.35 As described by Queraltò et al., such a mechanism leads to a small and scattered fraction of remnant polycrystalline grains on the surface, as observed in the present work.35 A viable strategy for eliminating such grains would be to use rapid thermal annealing during the pyrolysis step to increase the onset temperature of nucleation, which would favor preferred heterogeneous nucleation.
Consistent with this interpretation, we find that 10 min annealing at a lower temperature of 625 °C leads to an incomplete transformation, resulting in mixed domains of polycrystalline and epitaxial material (Fig. S16a†). In contrast, annealing at a higher temperature of 675 °C yields epitaxial films, though local pitting is observed (Fig. S16b†), the precise origin of which is not yet clear. Thus, 650 °C was chosen as the optimal annealing temperature for forming closed epitaxial films.
For all films, Tauc analysis yields an indirect band gap of 2.53 eV and a direct transition near 2.72 eV (Fig. S3†). Although slightly smaller direct band gaps have been previously reported for some epi-BiVO4 films,9,12,14–16 we find no significant difference between poly- and epi-films. Low and nearly identical Urbach energies are observed for all samples, with values of 65 meV for polycrystalline films on fused silica, 62 meV for the textured polycrystalline, and 58 meV for the epitaxial films on YSZ. However, epitaxial material exhibits reduced sub-band gap optical absorption relative to polycrystalline material, indicating moderate reduction of midgap states in the material. The overall optical quality of all films produced via metalorganic decomposition is found to be high, comparing favourably with prior reports6,36.
The availability of planar and compact epitaxial films offers the intriguing possibility to also investigate anisotropic material properties. Of particular relevance for solar energy conversion is the strong optical anisotropy of BiVO4, which has been computationally predicted37–39 but only partially investigated experimentally.31,40 Here, we performed variable-angle spectroscopic ellipsometry (VASE) to quantify the optical constants and compare results from epitaxial films on YSZ with polycrystalline material possessing random grain orientations. We note that a true anisotropic model cannot be reliably established with such a thin film given the weak contribution from out-of-plane oscillators and the large number of fitting parameters, which lead to non-physical parameter correlation. Therefore, in the present work we approximate the system using an isotropic general oscillator model to extract the optical constants of epi-BiVO4 and compare it to poly-BiVO4 with randomly oriented grains, as shown in Fig. 3b and c. The corresponding real and imaginary components of the dielectric function are presented in Fig. S4† and a detailed description of the optical model can be found in the Methods section. Importantly, we find a strong enhancement of the optical constants for the electric field E ‖ [100] and E ‖ [001] compared to E ‖ [010]. This result is consistent with prior DFT calculations, which predict band edge O 2p → V 3d dipole excitations associated with the fundamental band gap to be forbidden along the b-axis, but allowed along the a- and c-axes.38 To the best of our knowledge, this increase of both optical constants near the band edge due to anisotropy has not been previously reported. This significant increase of the extinction coefficient at the band edge for epi-BiVO4 (010) compared to polycrystalline material suggests a preferred BiVO4 (010) orientation for efficient light harvesting. Furthermore, this observation explains the different spectral shapes for band edge absorption that have been reported in the literature, which at times include a shoulder or peaked shape near 2.8 eV; films with different texture will naturally exhibit different above-gap absorption spectra due to relative sampling of the anisotropic dielectric function. This conclusion is supported by VASE measurements of textured poly-BiVO4/YSZ, which exhibit intermediate optical constants between ideally oriented epi-BiVO4 and randomly oriented poly-BiVO4/fused silica (Fig. S4†).
![]() | (1) |
![]() | ||
Fig. 4 Temperature-dependent in-plane conductivity of epi-BiVO4/YSZ. Conductivity, σ, plotted as ln(σT) vs. 103/T. The dashed line represents the fit to eqn (1), yielding a charge transport activation energy, Eσ, of 570 meV. In the inset, the conductivity is plotted against the temperature in a semi-log plot. |
The resulting charge transport activation barrier, Eσ, of 570 meV and the absolute conductivity (inset Fig. 4) are in excellent agreement with the previously reported values obtained from undoped BiVO4 on YSZ deposited via PLD.11 Interestingly, the films deposited by Zhang and co-workers were reported to comprise a single domain, whereas our films are characterized by twinning with domain boundaries. The finding of nearly identical transport barriers for both types of films suggests that in-plane transport is not significantly impacted by the presence of domain boundaries. This is surprising given that domain boundaries for epi-BiVO4 on indium-doped tin oxide (ITO) were found to act as high-conductivity channels for out-of-plane electrical transport,11 pointing to the need for improved understanding of planar defects in this material. Nevertheless, the electrical quality of the films is found to be commensurate with state-of-the-art material fabricated by PLD, thus confirming the applicability of solution-derived BiVO4 for fundamental studies of transport mechanisms.
Footnote |
† Electronic supplementary information (ESI) available: Additional XRD, contact angle, Raman, XPS, AFM, pole figures, 2D reciprocal space maps, and ellipsometry data. See DOI: 10.1039/d1ta10732a |
This journal is © The Royal Society of Chemistry 2022 |